Learning Objective 1: The learner will be able to appreciate the importance and public health consequences of accurate prediction of daily visits to a hospital emergency department
Learning Objective 2: The learner will be able to implement a new approach that predicts daily number of emergency department visits
Methods: Data of patient visits to the Medical Center ED were collected retrospectively throughout 1/1/2007-31/12/2011 (training set - TS) and 1/1-2/10/2012 (validation set - VS). A linear regression forecasting model for DPN comprised calendar-dependent data and the number of visits during morning hours in a given day.
Results: The mean DPN was: 559.3±94.2 (TS) and 615.1±101.8 (VS). DPNs differed according to the study years and weekdays (p<0.001 for each), yet not the months. Throughout two consecutive morning hours (8am-10am) a mean of 9.1±1.8% of the DPN occurred. The latter correlated with the DPN (r=0.79, p<0.001) and had an incremental prediction ability over the model based on calendar-dependent variables (year and weekday) both in the TS (R2 of 0.833 vs. 0.782) and the VS (R2 of 0.813 vs. 0.728) (p<0.001 for each). The MAPE decreased from 5.98 to 5.38 (TS) and from 6.57 to 5.55 (VS) (p<0.001 for each).
Conclusion: The number of ED visits during two specific morning hours is a robust predictor of DPN in a given day and has incremental prediction over calendar data alone. Accurate forecasting of DPN can help in optimizing hospital resources, avoiding overcrowding, and improve a healthcare quality.